Time averaged Reynolds stress tensor components in SU2 URANS

[kandevivek]

Hi,
TL;DR: I need mean RST components around flatback airfoil from SST URANS simulations, the output file does contain them but it seems to be very low everywhere except in the wake. I want to know if the output is correct or is there any other way to find Time averaged RST components/ TKE.

I am working on incompressible SST URANS simulations of flatback airfoils using SU2. I want to obtain the mean Reynolds stress tensor components~<u'u'>, <u'v'>.. . SU2 output does have these values, following is from the output from dry run:

|UUPRIME             |TIME_AVERAGE         |Mean Reynolds-stress component u'u'                     |
|VVPRIME             |TIME_AVERAGE         |Mean Reynolds-stress component v'v'                     |
|UVPRIME             |TIME_AVERAGE         |Mean Reynolds-stress component u'v'                     |

But the problem is when I plot these the contour shows non-zero values only in the wake, while in the boundary layer of airfoil it shows close to zero value. However the instantaneous turbulent kinetic energy shows non zero values even in the boundary layer as we would expect from TBL. So I am wondering if there is some issue with the calculation of Time average Reynolds stress components either from SU2 or if there is some issue with my setup for time averaging.
When I looked into the SU2 file CFlowOutput.cpp, it shows the following calculation for the mean Reynolds stress:

  const su2double umean  = GetVolumeOutputValue("MEAN_VELOCITY-X", iPoint);
  const su2double uumean = GetVolumeOutputValue("RMS_U", iPoint);
  const su2double vmean  = GetVolumeOutputValue("MEAN_VELOCITY-Y", iPoint);
  const su2double vvmean = GetVolumeOutputValue("RMS_V", iPoint);
  const su2double uvmean = GetVolumeOutputValue("RMS_UV", iPoint);
  const su2double pmean  = GetVolumeOutputValue("MEAN_PRESSURE", iPoint);
  const su2double ppmean = GetVolumeOutputValue("RMS_P", iPoint);

  SetVolumeOutputValue("UUPRIME", iPoint, -(umean*umean - uumean));
  SetVolumeOutputValue("VVPRIME", iPoint, -(vmean*vmean - vvmean));
  SetVolumeOutputValue("UVPRIME", iPoint, -(umean*vmean - uvmean));
  SetVolumeOutputValue("PPRIME",  iPoint, -(pmean*pmean - ppmean));

Since it is calculating Reynolds stress from mean and rms velocities, is there any chance it is missing the unresolved/ modelled turbulent stress components? I am not sure how rms[u] is being calculated I am assuming it doesn't have the unresolved fluctuation components.

Is there any other way to obtain either the time mean RST components or the time mean turbulent kinetic energy distribution through which I can approximately calculate the time averaged RST.

[sun5k]

Hi, @kandevivek.

The k-ω SST model is an eddy-viscosity model, meaning it uses turbulent viscosity instead of directly solving for the Reynolds stresses.

[kandevivek]

Hi,
Yes, I understand that. I was only expecting an estimate of RST components, not the direct solution, particularly their time averaged values. While I can use the Boussinesq eddy viscosity relation to find the instantaneous RST components using the eddy viscosity from the SST solver, I need the time averaged RST components (time averaged over the duration of simulation)- just an indirect estimate not direct values. Even if I could get the time averaged TKE (turbulent kinetic energy) that would be helpful because I can estimate the time averaged RST components from TKE through correlations in literature.

[pcarruscag]

Since you already found the code that does time averages, I suggest you modify it to average TKE, otherwise you can do it via the python wrapper.
The velocity fluctuations in any kind of URANS are not representative of the RST

[kandevivek]

Ok so the u' and v' in URANS solution output are not to be confused with turbulent fluctuations. Thanks for your response. The code I pasted here is from the GitHub page. I will checkout the python wrapper option.